Towards Easy and Realistic Network Infrastructure Testing for Large-scale Machine Learning

Fuente: arXiv
Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Yoo, Jinsun, Lao, ChonLam, Cao, Lianjie, Lantz, Bob, Yu, Minlan, Krishna, Tushar, Sharma, Puneet
Format: Preprint
Veröffentlicht: 2025
Schlagworte:
Online-Zugang:
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
_version_ 1866915265883668480
author Yoo, Jinsun
Lao, ChonLam
Cao, Lianjie
Lantz, Bob
Yu, Minlan
Krishna, Tushar
Sharma, Puneet
author_facet Yoo, Jinsun
Lao, ChonLam
Cao, Lianjie
Lantz, Bob
Yu, Minlan
Krishna, Tushar
Sharma, Puneet
contents This paper lays the foundation for Genie, a testing framework that captures the impact of real hardware network behavior on ML workload performance, without requiring expensive GPUs. Genie uses CPU-initiated traffic over a hardware testbed to emulate GPU to GPU communication, and adapts the ASTRA-sim simulator to model interaction between the network and the ML workload.
format Preprint
id arxiv_https___arxiv_org_abs_2504_20854
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Towards Easy and Realistic Network Infrastructure Testing for Large-scale Machine Learning
Yoo, Jinsun
Lao, ChonLam
Cao, Lianjie
Lantz, Bob
Yu, Minlan
Krishna, Tushar
Sharma, Puneet
Networking and Internet Architecture
Artificial Intelligence
Distributed, Parallel, and Cluster Computing
Systems and Control
This paper lays the foundation for Genie, a testing framework that captures the impact of real hardware network behavior on ML workload performance, without requiring expensive GPUs. Genie uses CPU-initiated traffic over a hardware testbed to emulate GPU to GPU communication, and adapts the ASTRA-sim simulator to model interaction between the network and the ML workload.
title Towards Easy and Realistic Network Infrastructure Testing for Large-scale Machine Learning
topic Networking and Internet Architecture
Artificial Intelligence
Distributed, Parallel, and Cluster Computing
Systems and Control
url https://arxiv.org/abs/2504.20854